Automatic Optical Inspection Using Multi-Model Neural Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automatic defect detection in manufacturing faces challenges in maintaining a low escape rate of defective products and minimizing false alarms, which affects production efficiency and product quality.
Innovation Solution
A novel framework for automatic optical inspection using multiple machine learning-based neural network models, including classification, segmentation, and anomaly detection modules, to classify images of products and determine their qualification status, thereby balancing escape and false alarm rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a single classification model is used for defect detection, then the escape rate can be maintained at a low level, but the false alarm rate increases significantly
Solution Approach 1:
The patent segments the defect detection task into three distinct neural network models: a classification model for initial defect identification, a segmentation model for precise defect boundary detection, and an anomaly detection model for verifying abnormal patterns. This segmentation allows each model to specialize in specific aspects of defect detection, thereby maintaining low escape rates while reducing false alarms through coordinated multi-model validation
Solution Approach 2:
The segmentation model serves as an intermediary between the classification model and the anomaly detection model. It processes the output of the classification model and provides refined defect information to the anomaly detection model, acting as a mediator that filters and refines detection results to reduce false alarms while maintaining detection accuracy
2Measurement precision
If multiple neural network models are used for inspection, then false alarm rate decreases, but device complexity increases
Solution Approach 1:
The detection framework is segmented into three specialized neural network models, each handling specific aspects of defect detection. This segmentation distributes the computational complexity across multiple focused models rather than requiring one overly complex model, making the system more manageable and efficient
Solution Approach 2:
The multi-model framework provides universal functionality by handling different types of defects and inspection scenarios through a coordinated system of specialized models. The classification model handles initial screening, the segmentation model handles precise measurement, and the anomaly detection model handles verification, creating a versatile inspection system
Data Source
AI summary
A method for automatic optical inspection includes (i) receiving an image of an object, (ii) classifying the image of the object as one of a plurality of categories by a classification model, (iii) determining a label for the image of the object as being qualified if the category obtained by the classification model is a first category, and (iv) performing defect measurement for the image by a segmentation model and determining a label for the image as being qualified or unqualified based on the defect measurement obtained by the segmentation model if the category obtained by the classification model is a second category.